EXCLUSIVE: Meet Tesla Bot’s Biggest Competitor | Brett Adcock Figure

In this talk

1. 3:52

   BMW deployment and next customers

   Adcock says Figure 2 completed a multiweek pilot at BMW’s Spartanburg site and stayed operational during the trial. Figure planned to place robots there permanently in 2025, with BMW paying for them; the goal was to make useful work happen daily and expand to more robots and tasks. Figure also aimed to introduce the robot to multiple customers early the following year.

2. 8:50

   Building at higher volume

   Adcock says Figure is building its own manufacturing facility and hiring people to design production lines, while acknowledging that high-rate humanoid production remains unsolved. He compares the manufacturing challenge with cars and phones, and says the company will make some components itself while using outside suppliers for others. Figure intends to control final assembly, testing, validation, and delivery.

3. 31:21

   Reliability, safety, and useful work

   Adcock says scaling requires long stretches of operation without human intervention, along with safety and enough output to meet customers’ business needs. In the lab, a person still helps in each work cell; Figure hopes to reduce that to one person overseeing three cells and eventually none. He describes adapting robots to varied tasks and work environments as a major challenge.

4. 25:02

   AI strategy and robot capabilities

   Adcock argues that factories offer a more controlled starting point than homes, where tasks and safety demands vary more. He says Figure had Microsoft support for H100 compute and was working with OpenAI on models, while remaining able to train without being constrained by compute or data at that time. Robots were designed to keep working offline, using onboard GPUs for fast actions and larger remote models when tasks allow more delay.

5. 46:55

   Designing the whole product

   Adcock distinguishes building a usable robot product from optimizing a single technology, saying the design must balance hardware, software, cost, reliability, manufacturing, and safety. Figure 1 helped develop AI and controls, while Figure 2 was designed to support a fuller range of humanlike work. He says the company makes tradeoffs across the system and iterates as it learns from each design.